I probably wouldn't rely on a transcription engine - the ones I'm familiar with require the user to read a long passage of pre-determined text so that the program can acclimate to their voice. Your therapist would certainly wonder why you'd ask her to do that in your session. Programs that don't require this sort of acclimatization to a user's specific accent and voice are notorious for misinterpreting what they hear. For example, I had a coworker whose iPhone texted her guy friend "thanks for the sex" instead of "thanks for lunch" when using speech to text.